RhoJ Regulates Melanoma Chemoresistance by Suppressing Pathways That Sense DNA Damage

Melanoma, an aggressive and fatal malignancy resistant to current therapies, has increased in incidence and mortality over the last several decades (1). The DNA alkylating agents dacarbazine or temozolomide are still used to treat metastatic melanoma, but only 16% of patients respond with no improvement in overall survival (2). Melanomas are resistant to a spectrum of chemotherapies including cisplatin, paclitaxel, docetaxel, combination chemotherapy, immunotherapy, and even newly developed BRAF inhibitors (3, 4). Melanoma cells acquire the ability to invade adjacent tissues and resist chemotherapy early during their evolution (4), further underscoring the importance of developing more effective treatments for this tumor.

Melanoma cells upregulate multiple pathways that allow them to be intrinsically resistant to apoptosis (4)—this includes the activation of anti-apoptotic factors [inhibitor of apoptosis protein (IAP) family, FLIP], the downregulation of pro-apoptotic genes (APAF-1, BAD, BIM), and the activation of prosurvival pathways (NF-κB, AKT). Despite the fact that only 10% to 20% of melanomas contain p53 mutations, most melanoma tumors contain additional mechanisms to suppress the function of p53, a central regulator of chemoresponsiveness (5). Transcriptional (6) and enzymatic regulators of melanogenesis (7) also modulate melanoma chemoresponsiveness. While extensive studies have identified multiple pathways that control melanoma chemoresistance, this information has not yet led to the development of more effective regimens to treat melanoma.

Synthetic lethal functional genomics screening is an emerging strategy to identify drug targets for the rational design of synergistic agents with selective toxicity toward cancer cells (8, 9). A gene and a drug have a synthetic lethal relationship if mutation or depletion of that gene sensitizes cells to sublethal concentrations of a drug (8, 9). Synthetic lethal screening has been used to identify genes that regulate lung cancer, cervical cancer, and breast cancer chemoresistance (9–11). In this study, we use a systems-level screening approach to identify regulators of melanoma chemoresistance with the goal of discovering pathways that could be the molecular targets of new synergistic chemotherapy regimens. This screen identified RhoJ, a CDC42 homologue that regulates endothelial cell migration and angiogenesis (12), as a novel regulator of melanoma cell chemoresponsiveness. We find that RhoJ activates PAK1 in response to DNA damage, which then suppresses the ability of ATR to activate its downstream effectors Chk1 and ATF2. ATR suppression ultimately results in decreased DNA damage–induced apoptosis and the increased expression of prosurvival genes. Taken together, these studies uncover a new signaling pathway that coordinately regulates survival and chemoresistance.

MNT-1 cells were a gift of M. Marks (University of Pennsylvania, Philadelphia, PA) and were cultured as described (13). C8161 melanoma cells were obtained from Frank Meyskens (University of California at Irvine, Irvine, CA). SK-Mel-28 melanoma cells were obtained from the American Type Culture Collection. SK-Mel-28 and C8161 melanoma cells were cultured in RPMI medium with 10% FBS. Darkly pigmented normal human melanocytes were purchased from Cascade Biologics and cultured as recommended by the manufacturer. The genome-wide siRNA library used in these studies and the transfection protocols were previously described (13). Dacarbazine and cisplatin were purchased from Sigma and IPA-3 was purchased from Tocris Biosciences. Cell line verification was conducted by Powerplex genotyping before use.

The antibodies for cleaved PARP, β-actin, p-Ser345-Chk1, Chk1, p-Ser20-p53, p73, phospho-Ser343-NBS1, NBS1, tubulin, ATF2, PLK1, PAK1, and phospho-PAK1,3 (Ser199/204)/PAK2 (Ser192/197) were purchased from Cell Signaling Technology. RhoJ (monoclonal from Abnova and polyclonal from Sigma), p53 (rabbit polyclonal, Santa Cruz Biotechnology), Sox10 (goat polyclonal, Santa Cruz Inc.), phospho-Ser490/498-ATF2 (Thermo Scientific), and p-Thr68-Chk2 and claspin (Abcam) were purchased as indicated.

High-throughput transfection was conducted as described (13). MNT-1 melanoma cells in plates were incubated for 72 hours at 37°C/5% CO₂, and light-activated dacarbazine (14) was added to make a final concentration of 0.1 mg/mL. After an additional 48 hours of drug incubation, a CellTiter-Glo Reagent (CTG; Promega) was delivered to each well, and the luminescence values for each well were determined. All SMART-pooled siRNAs used in this study were purchased from Dharmacon Inc.

SK-Mel-28 melanoma cells were transfected with 50 nmol/L pooled siRNAs. Forty-eight hours after transfection, cells were incubated in the presence and absence of 10 μmol/L cisplatin. Seventy-two hours after cisplatin treatment, treated cells were washed with PBS and incubated in fresh media for an additional 5 days to allow colony regrowth. Relative cell numbers were quantified using a sulforhodamine B assay (Sigma).

For the cell-cycle analysis, melanoma cells were trypsinized and washed with PBS and then were fixed in 70% ethanol. Cells were washed with PBS after fixation and were stained in a PBS solution containing 40 μg/mL propidium iodide (PI), 0.1 mg/mL RNaseA, 0.1% Triton-X 100. For the apoptosis assays, cells were trypsinized, washed with PBS, and stained with Alexa Fluor 488-Annexin V/PI according to the manufacturer's protocol (V13245; Invitrogen). After 15-minute staining, cells were subjected to flow cytometric analysis (BD FACSCalibur), and the resulting data were analyzed using FlowJo. We used cells that were not treated with cisplatin to establish gating parameters. Cells in the early-stage apoptosis were defined as the PI-negative Annexin V–positive population, whereas late-stage apoptotic cells were defined as the PI-positive Annexin V–positive population.

SK-Mel-28 melanoma cells or melanocytes were plated on 96-well plates. Twenty-four hours after plating, cisplatin and IPA-3 were added as indicated. Forty-eight hours after drug treatment, a CellTiter-Glo kit (Promega) was used to quantify the number of surviving cells. The relative percentage of surviving cells in IPA-3–treated and vehicle-treated samples was measured to calculate the synergy between IPA-3 and cisplatin.

We used a previously described (9) Dharmacon siRNA library of 84,508 siRNAs corresponding to 4 unique siRNA duplexes, targeting each of the 21,127 unique human genes arrayed in a one-gene/one-well format (4 siRNAs toward each gene) on 96-well microtiter plates to identify genes that selectively supported melanoma cell survival in the presence of dacarbazine (DTIC). MNT-1 cells, a cell line with intermediate resistance to a previously used in vitro genotoxic stress (light-activated dacarbazine; ref. 14), were transfected with 35 nmol/L target siRNA using published protocols (13) and incubated in the presence and absence of an empirically determined sublethal dose of dacarbazine (Fig. 1A). An ATP-dependent luminescence cell viability assay (CellTiter-Glo) was used to quantify the impact of each individual siRNA or siRNA + dacarbazine on cell viability. To identify genes that selectively support cell survival in the presence of dacarbazine, we calculated normalized luminescence values from dacarbazine-treated and -untreated wells for each siRNA and used these values to generate cell viability ratios (Supplementary Table S1). siRNAs with low viability ratios (lower cell numbers in siRNA-transfected, dacarbazine-treated wells than in siRNA-transfected, carrier-treated wells) correspond to dacarbazine synthetic lethal genes. In addition to identifying 140 candidate genes that modulate melanoma chemoresponsiveness (Supplementary Fig. S1A, red dots), we identified an even larger number of siRNAs that appeared to selectively promote cell survival in the presence of dacarbazine (z-score > 4). Closer analysis of these high z-score siRNAs revealed that they all potently induced cell death in the absence of dacarbazine (Supplementary Fig. S1B), whereas the addition of dacarbazine appeared to promote the survival of a fraction of the siRNA-transfected cells. Interestingly, this set of genes was enriched in kinases including BRAF (Supplementary Fig. S1C), the target of new melanoma therapies (15). While BRAF siRNAs potently inhibited ATP accumulation in MNT-1 cells (Supplementary Fig. S1B, red bar), the addition of either dacarbazine or cisplatin slightly inhibited the impact of BRAF depletion on ATP accumulation (Supplementary Fig. S1D). Recent clinical trials have shown that the combination of dacarbazine and BRAF inhibitors offers no improvement in survival when compared with patients treated with dacarbazine alone (16), suggesting that DNA damage agents and kinase inhibitors are not necessarily synergistic in the clinical setting. These findings reinforce the importance of establishing a mechanistic rationale for combination chemotherapies before the initiation of clinical trials.

Dacarbazine induces the methylation of guanine bases, resulting in base mispairings that are ultimately repaired by the DNA mismatch repair system (17). A dacarbazine-based synthetic lethal screening approach would be predicted to identify candidate genes that control mismatch repair while also identifying other candidate genes that suppress the cellular response to DNA damage. To identify the subset of candidates that suppress the DNA damage response (DDR), we next sought to identify siRNAs that sensitized melanoma cells to both dacarbazine and cisplatin, an agent that induces the formation of more bulky DNA adducts, which can be repaired by several different DNA repair mechanisms (17). Initial studies eliminated false-positives present in our candidate list (Supplementary Fig. S2A) and used a pool deconvolution approach to confirm that the observed dacarbazine sensitization phenotypes for our top 12 hits were not a result of RNA interference off-target effects using well-established criteria (Supplementary Fig. S2B; refs. 9, 13, 18). The top 12 siRNAs identified in the screen also modulated dacarbazine resistance in another melanoma cell line (SK-Mel-28 melanoma cells; Supplementary Fig. S2C). In addition, we validated that the siRNAs used in the study effectively inhibited the expression of their target genes (Supplementary Fig. S4A). Dose–response studies revealed that 7 of the 12 identified siRNAs potently sensitize SK-Mel-28 melanoma cells to cisplatin at low doses (RND2, RhoJ, SMARCE1, PPP1R12C, BLM, PMS2, and ARL4A; Fig. 1B). As predicted, 2 regulators of DNA damage resistance identified in the screen were DNA repair genes (PMS2 and BLM) known to repair damage induced by cisplatin (Table 1; refs. 19, 20). Two other genes (SMARCE1 and PPP1R12C) regulate essential processes that are required for DNA replication: chromatin remodeling (21) and mitosis (22), respectively. Three of the 7 validated genes were Ras superfamily GTPases (RhoJ, Arl4a, Rnd2), implicating a novel role for Ras superfamily GTPases in the DDR. Colony formation assays revealed that depletion of RhoJ, RND2, BLM, PPP1R12C, and SMARCE1 had profound effects on the proliferation/survival of cisplatin-treated SK-Mel-28 cells (Fig. 1B). Once we had identified the subset of candidate genes that potently regulate cell proliferation/survival, we next measured the accumulation of apoptosis markers (cleaved PARP) in siRNA-transfected, drug-treated cells to identify which of these siRNAs sensitize melanoma cells to cisplatin-induced apoptosis. While low-dose cisplatin treatment induced baseline levels of PARP cleavage in MNT-1 and SK-Mel-28 cells treated with control siRNAs (Fig. 1C), low-dose cisplatin induced significantly more PARP cleavage in MNT-1 and SK-Mel-28 cells treated with RhoJ, RND2, ARL4A, PMS2, and BLM siRNAs (Fig. 1C). Of interest, one of the genes that sensitized cells to cisplatin in both cell lines was ARL4A, an Arf-like GTPase that is genetically amplified in melanoma (23). The gene that most potently sensitized cells to cisplatin-induced death was RhoJ. Retrospective analysis of previously published microarray datasets revealed that RhoJ is overexpressed in metastatic melanoma (Supplementary Table S2; ref. 24), implicating a role for RhoJ in modulating chemoresponsiveness in human tumors. RhoJ siRNA efficiently inhibited RhoJ expression at both the RNA and protein level (Supplementary Fig. S4B). Moreover, RhoJ depletion potently sensitized MNT-1 cells (Fig. 1E) and SK-Mel-28 cells (Supplementary Fig. S3A and S3B) to cisplatin-induced apoptosis as measured by the accumulation of Annexin V–positive cells. In contrast, RhoJ overexpression enhanced melanoma chemoresistance (Fig. 1F), suppressed cisplatin-induced PARP cleavage (Fig. 1G), and also inhibited cisplatin-induced apoptosis (Fig. 1H). Taken together, these results identify RhoJ as a suppressor of DNA damage–induced apoptosis.

While published studies have revealed that Rho and CDC42 signaling pathways regulate melanoma invasion (25), these GTPases were not known to regulate chemoresistance. To determine how RhoJ regulates melanoma chemoresistance, we sought to identify downstream pathways activated by RhoJ and determined how these pathways modulate the DDR. RhoJ is a CDC42 homologue, a class of GTPases that can bind and activate group I Pak kinases that contain Pak autoinhibitory domains (26). We used a group I Pak inhibitor (IPA-3) to determine whether Pak inhibition sensitized melanoma cells to cisplatin-induced apoptosis. IPA-3 and cisplatin acted synergistically to inhibit ATP accumulation in melanoma cells (Fig. 2A). On the other hand, melanocytes were very sensitive to low-dose IPA-3 (Fig. 2B), and addition of cisplatin only slightly enhanced the inhibitory effect of IPA-3 on ATP accumulation (Fig. 2B). Next, we asked whether group I Pak kinases (PAK1, PAK2, or PAK3) are activated by RhoJ. Using an antibody that can recognize phospho-Ser199/204 PAK1/3 or Ser192/197PAK2, we observed that cisplatin treatment induced the accumulation of a band that corresponds to p-PAK1/3 in both MNT-1 and SK-Mel-28 melanoma cells (Fig. 2C). Cisplatin did not induce the accumulation of a band that corresponds to phospho-Pak2, which migrates at a different molecular weight, indicating that cisplatin activates either PAK1/3. Using antibodies specific for PAK1 and PAK3, we show that Pak3 expression is undetectable in the cells studied (Supplementary Fig. S4C), suggesting that cisplatin activates PAK1. Importantly, cisplatin induced the accumulation of phospho-PAK1 in a RhoJ-dependent manner (Fig. 2C), whereas Pak1 depletion sensitized both SK-Mel-28 and MNT-1 melanoma cells to cisplatin-induced PARP cleavage (Fig. 2D). These results are consistent with published studies that have shown that RhoJ can activate PAK1 (27) and indicate that RhoJ modulates melanoma chemoresponsiveness by activating PAK1. If this is true, then depletion of PAK1 in RhoJ-overexpressing cells should sensitize these cells to cisplatin-induced apoptosis. While RhoJ overexpression induced the phosphorylation of PAK1 and suppressed cisplatin-induced PARP cleavage (Fig. 2E), depletion of PAK1 in the context of RhoJ overexpression inhibited phospho-PAK1 accumulation and restored the level of PARP cleavage observed in vector-expressing cells (Fig. 2E), consistent with the contention that PAK1 is activated by RhoJ to deflect DNA damage–induced death. While we cannot exclude that RhoJ can activate other group I Paks in other melanoma cell lines when they are expressed, our results indicate that RhoJ activates PAK1 to modulate melanoma chemoresponsiveness in the cell lines studied.

Once we had identified a potential downstream mediator of RhoJ signaling, we next sought to examine whether RhoJ/Pak1 modulates the DDR. Our initial screen revealed that RhoJ depletion sensitized cells to dacarbazine, an agent that generates DNA mismatches that are known to activate the ATR kinase [ref. 28 (#10926)]. Once activated, ATR phosphorylates multiple different targets, including H2AX and Chk1, which can then mark the sites of stalled replication forks or induce cell-cycle arrest/apoptosis, respectively (17). There is some inherent redundancy in this system, as ATM, a kinase that is induced upon double-strand break formation, can also phosphorylate H2AX when ATR is not active (29). Cisplatin induced similar amounts of p-H2Ax accumulation in RhoJ-depleted or control siRNA–treated cells (Supplementary Fig. S5A). RhoJ depletion itself did not induce DNA damage and cisplatin induced a similar amount of DNA damage in RhoJ-depleted or control siRNA–transfected cells (Supplementary Fig. S5B). Taken together, these observations indicate that RhoJ does not inhibit DNA damage from occurring.

Once we had determined that RhoJ does not prevent DNA damage, we sought to determine whether RhoJ/PAK1 modulates the cellular response to DNA damage. Upon DNA damage, ATR localizes to chromatin where it is activated (30). ATR then phosphorylates Chk1 on Ser345, which can initiate cell-cycle arrest or induce apoptosis (30). RhoJ depletion when coupled with cisplatin treatment induced the selective accumulation of Chk1 phosphorylated at its ATR-dependent sites (Fig. 3A), indicating that RhoJ modulates ATR activity. Surprisingly, pChk2, a kinase that is phosphorylated specifically by ATM (28), accumulates in untreated cells but not in cisplatin-treated cells (Fig. 3A), suggesting that RhoJ modulates ATR and not ATM activity. PAK1 depletion or IPA-3 treatment when coupled with cisplatin treatment also induced the accumulation of pChk1 (Fig. 3A), indicating that RhoJ/PAK1 normally suppress Chk1 activation. Finally, RhoJ overexpression suppressed cisplatin-induced accumulation of p-Ser345-Chk1 (Fig. 3B), showing that RhoJ and Pak1 inhibits the ability of ATR to phosphorylate Chk1 (30).

Once we had shown that RhoJ suppresses ATR activation, we next asked whether RhoJ regulates cisplatin-induced apoptosis in an ATR-dependent manner. While RhoJ depletion when coupled with cisplatin induced the accumulation of cleaved PARP, co-depletion of ATR and RhoJ suppressed cisplatin-induced cleaved PARP accumulation in both p53 wild-type (C8161) and p53-mutant (SK-Mel-28) cells (Fig. 3C). Similarly, depletion of RhoJ or ATR alone when coupled with cisplatin treatment induced the accumulation of Annexin V–positive cells (Fig. 3D, top), whereas co-depletion of ATR and RhoJ suppressed the cisplatin-induced accumulation of Annexin V–positive cells (Fig. 3D, bottom). Depletion of Chk1 alone was highly toxic, so we were unable to test whether RhoJ modulates chemoresistance in a Chk1-dependent manner (data not shown). Nonetheless, our observations indicate that RhoJ suppresses the DDR by an ATR-dependent mechanism.

Next, we sought to determine how RhoJ/Pak1 uncouples ATR from Chk1. Claspin, a scaffold that couples ATR to its downstream effector Chk1 (31), can be phosphorylated by the PAK1 target PLK1 (32). Phosphorylated claspin is subsequently degraded by the ubiquitin–proteasome pathway (31). We found that RhoJ overexpression suppressed claspin accumulation, whereas RhoJ depletion stimulated claspin accumulation (Fig. 3E), suggesting that RhoJ modulates ATR signaling by modulating claspin accumulation. Similarly, PAK1 depletion modulated claspin accumulation both in the absence and in the presence of cisplatin (Fig. 3F). RhoJ overexpression induced the accumulation of Plk1, the kinase that phosphorylates claspin and induces its degradation. PAK1 depletion suppressed the accumulation of Plk1 induced by RhoJ overexpression (Fig. 3G), consistent with a role for PAK1 in activating PLK1. Taken together, our results indicate that RhoJ modulates the DDR by suppressing the ability of ATR to phosphorylate its downstream effectors, including Chk1.

Published studies have indicated that Chk1 primarily functions to induce cell-cycle arrest (33), although it can also initiate apoptosis by phosphorylating p53 (34). Depletion of RhoJ suppressed proliferation (Fig. 4A, representative images in Supplementary Fig. S5D) and enhanced S-phase arrest (Fig. 4B) in the presence but not in the absence of cisplatin in both p53 wild-type and mutant cells, indicating that RhoJ suppressed Chk1-dependent cell-cycle arrest. As Chk1 can also induce apoptosis by activating p53, we next sought to determine whether RhoJ regulates cisplatin-induced apoptosis in a p53-dependent manner. PAK1 and RhoJ depletion induced the selective accumulation of p53 and p53 phosphorylated at Ser20 (a Chk1-dependent phosphorylation site; ref. 34) in p53 wild-type cells (Fig. 4C). In contrast, PAK1 or RhoJ depletion did not induce the accumulation of p53 (Fig. 4D) or p-Ser20-p53 (data not shown) in p53 mutant SK-Mel-28 melanoma cells. Co-depletion of p53 and RhoJ suppressed cisplatin-induced PARP cleavage (Fig. 4E) and cisplatin-induced Annexin V accumulation in p53 wild-type cells (Fig. 4F). While co-depletion of p53 and RhoJ suppressed cisplatin-induced PARP cleavage in p53-mutant cells (SK-Mel-28; Fig. 4E), it did not inhibit Annexin V accumulation in these cells (Supplementary Fig. S3C). Taken together, these results suggest that RhoJ regulates cisplatin-induced apoptosis in a p53-dependent manner in wild-type cells. In p53-mutant cells, the ability of RhoJ to modulate chemoresponsiveness is only partially dependent on p53, consistent with published studies indicating that mutant p53 can be partially functional (35).

Next, we sought to better understand how RhoJ regulates melanoma chemoresponsiveness in p53-mutant cells. In the absence of functional p53, Chk1 can modulate apoptosis by phosphorylating the p53 homologue p73 (36). Unfortunately, co-depletion of p73 and RhoJ did not inhibit cisplatin-induced Annexin V accumulation (Supplementary Fig. S3C). Still other studies have identified NBS1 as a protein that is a downstream target of ATR (37) that can modulate DNA damage–induced apoptosis (38). Co-depletion of NBS1 and RhoJ did suppress cisplatin-induced apoptosis in p53-mutant cells (Supplementary Fig. S3C). Moreover, cisplatin treatment induced the phosphorylation of NBS1 at its ATR-specific phosphorylation site only in RhoJ siRNA–treated cells (Supplementary Fig. S3D), indicating that RhoJ normally suppresses cisplatin-induced NBS1 phosphorylation. Taken together, these results suggest that RhoJ regulates melanoma chemoresistance by suppressing the ability of ATR to phosphorylate its downstream effectors, resulting in decreased activation of Chk1 and the decreased phosphorylation of Chk1 targets including p53. In p53-mutant cells, RhoJ can suppress cisplatin-induced apoptosis both by inhibiting the activation of hypofunctional p53 and by suppressing the ability of ATR to phosphorylate other effectors, such as NBS1. Interestingly, other targets identified in our screen did not modulate Chk1 activation but did modulate p53 accumulation (Supplementary Fig. S5C), reinforcing the concept that p53 is a central regulator of melanoma chemoresponsiveness.

To gain a better appreciation of molecular pathways regulated by RhoJ, we next sought to identify RhoJ-regulated genes and pathways using a microarray-based approach (see dataset, Supplementary Table S3). Gene set enrichment analysis (39) revealed that regulators of neural development were significantly up- or downregulated in cells expressing RhoJ short hairpin RNA (shRNA) when compared with cells expressing control shRNAs (Supplementary Fig. S6A). In particular, chronic RhoJ depletion potently inhibited the expression of Sox10 (>100-fold) as well as Sox10 target genes (Supplementary Fig. S6B). Chronic RhoJ depletion was sufficient to inhibit both Sox10 mRNA/protein accumulation in C8161 and SK-Mel-28 melanoma cells (Fig. 5A). While RhoJ depletion did not inhibit the accumulation of Sox10 mRNA in melanocytes, it did inhibit the accumulation of Sox10 protein in these cells (Fig. 5A). RhoJ overexpression induced the accumulation of Sox10 mRNA and protein in both SK-Mel-28 cells and normal melanocytes (Fig. 5B), indicating that RhoJ regulates the expression of Sox10.

Sox10, a known oncogene in melanomas, regulates the survival of neural crest precursor cells before lineage commitment in the developing embryo (40) and is a reliable marker for melanoma tumor–initiating cells (41). Recent studies have determined that Sox10 is required for melanoma formation in both mice and humans and determined that depletion of Sox10 alone is sufficient to induce apoptosis in melanoma cells (42). ATF2, a transcription factor that is activated by ATM/ATR signaling, has recently been shown to suppress Sox10 expression in melanoma cells (43). Depletion of ATR inhibited the phosphorylation of ATF2 at its ATR/ATM target sites while also inducing the accumulation of Sox10 (Fig. 5C), consistent with a role for ATR in modulating ATF2 function and repressing Sox10 expression. In contrast, RhoJ or Pak1 depletion or Pak1 inhibition resulted in the accumulation of ATF2 phosphorylated at its ATR target sites (Fig. 5D). PAK1 depletion or inhibition ultimately resulted in the decreased accumulation of Sox10 (Fig. 5E). Sox10 depletion inhibited melanoma cell survival (Supplementary Fig. S6C) and sensitized melanoma cells to cisplatin-induced PARP cleavage (Fig. 5F). In addition, cisplatin treatment inhibited Sox10 accumulation (Fig. 5G), indicating that ATR activation can suppress Sox10 expression. Intriguingly, Sox10 overexpression inhibited cisplatin-induced apoptosis in p53-mutant cells (Fig. 5H), and Sox10 overexpression partially mitigated the impact of Pak inhibitors on cisplatin sensitivity (Fig. 5I). These results indicate that Sox10 expression also promotes melanoma chemoresistance. In summary, our results indicate that RhoJ activates PAK1 in response to DNA damage. Activated Pak1 then suppresses claspin accumulation, which uncouples ATR from its downstream effectors Chk1 and ATF2. This ultimately results in the functional inactivation of p53 and the increased expression of prosurvival genes (Fig. 6).

In this study, we have used an unbiased genome-wide functional genomics approach to identify RhoJ and its downstream kinase PAK1 as novel regulators of response of the melanoma cell to DNA damage. Cisplatin activates PAK1 in a RhoJ-dependent fashion, indicating that RhoJ activates PAK1 to modulate melanoma chemoresistance. While RhoJ does not inhibit cisplatin-induced DNA damage, it does suppress the phosphorylation of the ATR downstream effectors Chk1 and ATF2. In p53 wild-type cells, RhoJ-induced ATR suppression leads to a functional inactivation of p53. In p53-mutant cells, RhoJ-induced ATR suppression results in the decreased phosphorylation of the ATR target NBS1 and the increased expression of Sox10, which can also suppress drug-induced apoptosis. Importantly, parallel studies have determined that PAK1 depletion or Pak kinase inhibition recapitulated all of the phenotypes observed upon RhoJ depletion, suggesting that RhoJ regulates all of these phenotypes, in part, by activating PAK1. Taken together, these studies identify the suppression of DNA damage sensing, as opposed to the commonly accepted mechanism of apoptosis execution, as a linchpin of chemoresistance in melanoma and nominate RhoJ and PAK1 as optimal therapeutic targets for the rational design of novel synergistic chemotherapy regimens.

Melanocytes in the epidermis produce melanin in response to UVB irradiation and transfer this pigment to keratinocytes, protecting them from UV-induced DNA damage (44). To accomplish this, melanocytes must possess mechanisms to resist DNA damage–induced apoptosis and also facilitate the expression of genes that regulate melanin production and survival (45). Of interest, our results suggest that RhoJ and PAK1 limit the response to DNA damage while concomitantly facilitating the expression of central regulators of melanogenesis. RhoJ/PAK1 suppress ATR activation, which can also be induced by UV, allowing the melanocyte to tolerate higher levels of DNA damage than other epidermal cells. ATR suppression in turn leads to the increased expression of Sox10, which can in turn regulate MITF, a transcription factor that controls the expression of multiple genes required for melanin production (46). Elevated expression of Sox10 is observed in melanoma tumors, and Sox10 is required for melanoma formation (42). Similarly, RhoJ is overexpressed in advanced melanomas as compared with primary melanomas (Supplementary Table S3). In addition, melanoma tumors that occur in UV-exposed skin are known to accumulate large numbers of mutations (47), suggesting they have an increased DNA damage tolerance. These observations suggest that in the normal setting, RhoJ/PAK1 activity allows melanocytes to tolerate limited amounts of DNA damage to facilitate the production of melanin in response to UV stress. When this pathway is activated in melanoma, the resulting tumor cells are profoundly resistant to DNA damage agents and have an increased DNA damage tolerance as evidenced by their accumulation of multiple mutations. Taken in context, our results indicate that melanoma cells are intrinsically chemoresistant because they activate cell-autonomous, lineage-selective pathways that blunt the responsiveness of the DNA damage surveillance machinery, thus allowing them to persist in the face of high levels of DNA damage. Development of agents that selectively disable RhoJ/PAK1 pathway activity may therefore represent an opportunity to resensitize melanoma tumors to conventional chemotherapeutic agents.

No potential conflicts of interest were disclosed.

Conception and design: H. Ho, M.A. White, A.K. Ganesan

Development of methodology: H. Ho, J. Aruri, M.A. White, A.K. Ganesan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Ho, H. Mehr, M.A. White, A.K. Ganesan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Ho, R. Kapadia, H. Mehr, M.A. White, A.K. Ganesan

Writing, review, and/or revision of the manuscript: H. Ho, M.A. White, A.K. Ganesan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Ho, A.K. Ganesan

Study supervision: H. Ho, A.K. Ganesan

Carrying different experiments for the data: R. Kapadia

The authors thank Stacie Loftus and Bill Pavan for their suggestions, Keith Hoek for his assistance with the RhoJ tumor expression data, and Kyoko Yokomori and Roberta Baronio for their advice about the COMET assay and p53 sequencing.

This work was supported by the NIH (1K08AR056001 to A.K. Ganesan and CA71443 to M.A. White), the UC Cancer Research Coordinating Committee, the American Cancer Society (121540-RSG-11-128-01-CSM to A.K. Ganesan), Outrun the Sun, Inc., and the Robert E. Welch Foundation (I-1414 to M.A. White). The microscopy studies were supported by a P41-RR01192. This work was also partially supported by UL1 RR031985 from the National Center for Research Resources (NCRR).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.